Skip to content
Cyntex Docs

Authoring with AI

1. Import the Cyntex skill → AI learns DSL syntax
2. Describe your need in conversation → AI outputs YAML
3. cyntex validate → local three-layer validation
4. cyntex apply → deploy to running Cyntex
5. (optional) MCP real-time monitoring of runtime status

Step 1: Help the AI Understand Cyntex

Provide the following URL to your AI:

https://docs.cyntex.io/llms.txt

Or paste the contents of llms.txt directly into the conversation. The AI will then understand Cyntex’s core concepts and DSL syntax.

Step 2: Describe Your Requirements

I have a MySQL database and I need to:
- Sync the users table to MongoDB in real time (CDC mode)
- Filter out records where deleted_at is not null
- Rename the user_id field to id


Please generate the corresponding .cyn.yml files

Step 3: Validate the Generated YAML

Terminal window
cyntex validate user-sync.cyn.yml

AI-generated YAML may have minor errors. validate gives precise error locations and explanations. Feed the errors back to the AI for correction:

validate error:
  Error: Unknown field 'write_strategy' at sync[0].options
  Did you mean 'write_mode'?


Please fix this field name

Best Practices

Tips for prompting AI:

  • Clearly state the type of source and target (MySQL CDC → MongoDB)
  • Specify which transforms are needed (filter, rename, enrichment)
  • Mention whether DDL changes need to be handled (ddl: apply)
  • Describe one pipeline at a time; split complex scenarios across multiple conversations

Validate before deploying:

cyntex validate is free — always run it before apply, especially for AI-generated content.

Use explain to understand fields:

Terminal window
cyntex explain pipeline.sync.options.write_mode
# Outputs field documentation, valid values, and defaults

Combining with MCP

After the Alpha phase, AI can not only generate YAML but also deploy and monitor directly via MCP:

User: Check the lag on the user-sync task
AI: [calls get_pipeline_status("user-sync")]
    Current lag: 1.2s, past 1-hour average: 0.8s — status normal